Human Classification Using Gait Features

The main purpose of the project is the body movement analysis using ICT tools that are cheap, non-intrusive and in real-time. We perform Gait analysis using skeletal data provided by the Microsoft Kinect sensor. Our approach shows that a limited set of behavioral features related to the movements of head, elbows and knees is a very effective tool for gait characterization and people recognition.

Kinect Gait Dataset:

For our experiments we built a dataset of gait samples, named KinectUNITO.
The dataset is available on request.

The purpose of this dataset is to provide data for testing and evaluating algorithms for people classification and recognition using Skeleton Data. We acquired the skeleton model of 20 subjects, 12 males and 8 females, using Kinect for Windows v1 and the Microsoft SDK. The Kinect sensor has been placed in a corridor, at about one meter from the ground and every subject has been asked to walk towards the camera in a natural manner. Every acquisition has been repeated 10 times and includes both the walk toward and far away from the camera, i.e. the frontal and the rear view respectively.
Therefore, we collected a total of 400 gait samples, 200 for the frontal view and 200 for the rear one.

The dataset is composed by 20 text file. Each file represents a subject and collects all his gait sequences, with the following order: FRONT, REAR, FRONT, REAR, and so on. No filter is applied to the data.

Here you can download a sample of the dataset.

If you are interested in our dataset please contact me ().

For more information about our results please refer to the following papers:
  • GIANARIA E., Marco Grangetto, Maurizio Lucenteforte, and Nello Balossino. Human classification using gait features. In 1st International Workshop on Biometrics, BIOMET 2014, volume 8897, pages 16-27, Cham - CHE, 23-24/06/ 2014. Springer Verlag. [ bib | DOI | http ]
  • GIANARIA E., N. Balossino, M. Grangetto, and M. Lucenteforte. Gait characterization using dynamic skeleton acquisition. In IEEE 15th International Workshop on Multimedia Signal Processing, pages 440-445, Newark - USA, 30/9/ 2013. IEEE. [ bib | DOI | http ]

Come back to the Projects list